Traffic Flow Prediction of Highway Based on Wavelet Relevance Vector Machine

摘要

Accurate prediction for traffic flow is the key to ensure the reasonable dispatching of traffic flow. In the study, wavelet relevance vector machine is proposed to predict traffic flow of highway. The traffic flow data of a certain highway road from November 27 to November 28 in 2002 are used to study the prediction performance of the proposed method. The traffic flow data are recorded in each interval 15 minutes, then, 192 traffic flow data are gained, where the first 142 traffic flow data are used to train the proposed prediction model and others are used to test the proposed prediction model. The comparison of the prediction precision of traffic flow between WRVM and SVM in the multi-step prediction to judge the ability of them. The prediction results based on the different dimension of input vector are given and the comparison of prediction results between WRVM and SVM based on 6 dimension input vector is given. By comparing the experimental results, we conclude that the prediction accuracies of traffic flow of WRVM are higher than those of SVM.